Abstract:
In the current social and economic conditions, academic and corporative practice
the prospects of some variation series alignment methods and the goodness-of-fit tests line
using are particularly defined by the available number of elements sets. Usually for the ultrasmall sample size which units number doesn't exceed twenty-five the problem of relation
estimation becomes complicated with the current "turbulent economy" features when the
stable trends duration of the analyzed sets often doesn't exceed several units. In these
conditions even the Romanovsky and Pearson fitting criterion using will be concerned with
the cases of discrepancy to its calculated and tabular values. During the pandemic period the
data sets could be even smaller than usual, so here we propose to use the modified scale n.a.
prof. V.A. Yadov which could estimate the correlation between the nature of sample
representativeness in a linguistic scale and its error size in the interval scale in percentage
points. In this case we offer to modify a scale with the specification of a linguistic variable
terms names: an error size up to three percent could be characterized as a "minimum
deviation"; from three percent to ten percent as the "usual deviations"; from ten percent to
twenty percent as "tentative estimation deviations"; from twenty percent to forty percent as
"evaluative deviations" and in the case of error more than forty percent as the "rough
deviations". With such approach the goodness-of-fit tests computed results have the special
sense also for the analyzed short statistical series.
